Experimental MCP proxy for filtering large tool lists using hybrid semantic and keyword retrieval
Single tool 'shutup__set_intent' has a basic description (47 chars) and minimal schema. The tool name lacks a clear action verb (set_ is weak for a filtering/intent-setting operation). Input schema is present but sparse: only one required parameter 'intent' with a generic string type and brief description (56 chars). No output schema documented. No error handling guidance. The tool appears to be a control/configuration mechanism rather than a primary agent tool, which limits its utility in a production agent context. Tool definition is visible in shutup/proxy.py but lacks depth expected of production-grade tools.
Set the current intent used by shutup-mcp to filter tools/list results.
Tool name 'shutup__set_intent' lacks a clear action verb. 'set_intent' is vague, does it filter tools, configure the proxy, or change agent behavior? Rename to 'filter_tools_by_intent' or 'configure_tool_filter' to clarify the action.
No output schema documented. LLMs cannot plan downstream actions or validate the response structure. Document what the tool returns (e.g., filtered tool list, confirmation, or status).
Parameter 'intent' description is generic (56 chars). Does not explain when to call this tool, what intent values are valid, or what happens after setting intent. Add examples and constraints.
No error handling guidance. What happens if intent is empty, malformed, or matches no tools? How should the LLM recover?
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 40 | 2026-07-28+ | v2 |
Tool is a proxy control mechanism, not a primary agent tool. Its role in the agent workflow is unclear. Documentation should explain when and why an agent would call this vs. relying on implicit filtering.